feature flag

A feature flag is a switch built into a program that lets developers turn a newly built feature on or off without rebuilding the whole application. Think of a building's circuit breaker panel: instead of cutting power to the entire building, you can flip off just one room's lights because the wiring is divided into separate circuits. In the same way, developers plant a hidden handle inside the code so they can quietly switch off just one feature.

Why does this handle matter? Because not every problem can be caught before launch. Some issues only surface after a feature goes live. With the handle in place, that one feature can be switched off quietly while everything else keeps running. Without it, the entire application would need to be rebuilt and redeployed, which stretches out the time between spotting a problem and actually fixing it.

That's why services that can show a new feature to only a portion of users to gauge reactions, or roll something back within hours when trouble hits, are usually the ones that built this kind of switch in advance. Services that launch features without this structure tend to struggle to respond quickly when something goes wrong.

The article notes that "without a structure to turn things on and off via feature flags, a feature can't be pulled within a day." It points out that Google was able to withdraw its Earth AI feature just a day after launch precisely because it had built this kind of switch in advance.

One common misunderstanding: having a feature flag doesn't mean problems never occur in the first place. It's a mechanism that allows a fast rollback once a problem appears — it doesn't replace pre-launch testing.

If a new feature has ever suddenly appeared in an app you use and then quietly disappeared a few days later, that's likely a trace of a feature flag being switched on and off. If a feature appeared or vanished without an app update, it was probably a switch being flipped rather than new code being deployed.